1 /*===------------- avx512vnniintrin.h - VNNI intrinsics ------------------===
2 *
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a copy
5 * of this software and associated documentation files (the "Software"), to deal
6 * in the Software without restriction, including without limitation the rights
7 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8 * copies of the Software, and to permit persons to whom the Software is
9 * furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20 * THE SOFTWARE.
21 *
22 *===-----------------------------------------------------------------------===
23 */
24 #ifndef __IMMINTRIN_H
25 #error "Never use <avx512vnniintrin.h> directly; include <immintrin.h> instead."
26 #endif
27
28 #ifndef __AVX512VNNIINTRIN_H
29 #define __AVX512VNNIINTRIN_H
30
31 /* Define the default attributes for the functions in this file. */
32 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512vnni"), __min_vector_width__(512)))
33
34
35 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_dpbusd_epi32(__m512i __S,__m512i __A,__m512i __B)36 _mm512_dpbusd_epi32(__m512i __S, __m512i __A, __m512i __B)
37 {
38 return (__m512i)__builtin_ia32_vpdpbusd512((__v16si)__S, (__v16si)__A,
39 (__v16si)__B);
40 }
41
42 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_dpbusd_epi32(__m512i __S,__mmask16 __U,__m512i __A,__m512i __B)43 _mm512_mask_dpbusd_epi32(__m512i __S, __mmask16 __U, __m512i __A, __m512i __B)
44 {
45 return (__m512i)__builtin_ia32_selectd_512(__U,
46 (__v16si)_mm512_dpbusd_epi32(__S, __A, __B),
47 (__v16si)__S);
48 }
49
50 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_dpbusd_epi32(__mmask16 __U,__m512i __S,__m512i __A,__m512i __B)51 _mm512_maskz_dpbusd_epi32(__mmask16 __U, __m512i __S, __m512i __A, __m512i __B)
52 {
53 return (__m512i)__builtin_ia32_selectd_512(__U,
54 (__v16si)_mm512_dpbusd_epi32(__S, __A, __B),
55 (__v16si)_mm512_setzero_si512());
56 }
57
58 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_dpbusds_epi32(__m512i __S,__m512i __A,__m512i __B)59 _mm512_dpbusds_epi32(__m512i __S, __m512i __A, __m512i __B)
60 {
61 return (__m512i)__builtin_ia32_vpdpbusds512((__v16si)__S, (__v16si)__A,
62 (__v16si)__B);
63 }
64
65 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_dpbusds_epi32(__m512i __S,__mmask16 __U,__m512i __A,__m512i __B)66 _mm512_mask_dpbusds_epi32(__m512i __S, __mmask16 __U, __m512i __A, __m512i __B)
67 {
68 return (__m512i)__builtin_ia32_selectd_512(__U,
69 (__v16si)_mm512_dpbusds_epi32(__S, __A, __B),
70 (__v16si)__S);
71 }
72
73 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_dpbusds_epi32(__mmask16 __U,__m512i __S,__m512i __A,__m512i __B)74 _mm512_maskz_dpbusds_epi32(__mmask16 __U, __m512i __S, __m512i __A, __m512i __B)
75 {
76 return (__m512i)__builtin_ia32_selectd_512(__U,
77 (__v16si)_mm512_dpbusds_epi32(__S, __A, __B),
78 (__v16si)_mm512_setzero_si512());
79 }
80
81 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_dpwssd_epi32(__m512i __S,__m512i __A,__m512i __B)82 _mm512_dpwssd_epi32(__m512i __S, __m512i __A, __m512i __B)
83 {
84 return (__m512i)__builtin_ia32_vpdpwssd512((__v16si)__S, (__v16si)__A,
85 (__v16si)__B);
86 }
87
88 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_dpwssd_epi32(__m512i __S,__mmask16 __U,__m512i __A,__m512i __B)89 _mm512_mask_dpwssd_epi32(__m512i __S, __mmask16 __U, __m512i __A, __m512i __B)
90 {
91 return (__m512i)__builtin_ia32_selectd_512(__U,
92 (__v16si)_mm512_dpwssd_epi32(__S, __A, __B),
93 (__v16si)__S);
94 }
95
96 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_dpwssd_epi32(__mmask16 __U,__m512i __S,__m512i __A,__m512i __B)97 _mm512_maskz_dpwssd_epi32(__mmask16 __U, __m512i __S, __m512i __A, __m512i __B)
98 {
99 return (__m512i)__builtin_ia32_selectd_512(__U,
100 (__v16si)_mm512_dpwssd_epi32(__S, __A, __B),
101 (__v16si)_mm512_setzero_si512());
102 }
103
104 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_dpwssds_epi32(__m512i __S,__m512i __A,__m512i __B)105 _mm512_dpwssds_epi32(__m512i __S, __m512i __A, __m512i __B)
106 {
107 return (__m512i)__builtin_ia32_vpdpwssds512((__v16si)__S, (__v16si)__A,
108 (__v16si)__B);
109 }
110
111 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_dpwssds_epi32(__m512i __S,__mmask16 __U,__m512i __A,__m512i __B)112 _mm512_mask_dpwssds_epi32(__m512i __S, __mmask16 __U, __m512i __A, __m512i __B)
113 {
114 return (__m512i)__builtin_ia32_selectd_512(__U,
115 (__v16si)_mm512_dpwssds_epi32(__S, __A, __B),
116 (__v16si)__S);
117 }
118
119 static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_dpwssds_epi32(__mmask16 __U,__m512i __S,__m512i __A,__m512i __B)120 _mm512_maskz_dpwssds_epi32(__mmask16 __U, __m512i __S, __m512i __A, __m512i __B)
121 {
122 return (__m512i)__builtin_ia32_selectd_512(__U,
123 (__v16si)_mm512_dpwssds_epi32(__S, __A, __B),
124 (__v16si)_mm512_setzero_si512());
125 }
126
127 #undef __DEFAULT_FN_ATTRS
128
129 #endif
130